Report

Reversion of antibiotic resistance in Mycobacterium tuberculosis by spiroisoxazoline SMARt-420

+ See all authors and affiliations

Science  17 Mar 2017:
Vol. 355, Issue 6330, pp. 1206-1211
DOI: 10.1126/science.aag1006

Countering TB prodrug resistance

The arsenal of antibiotics for treating tuberculosis (TB) contains many prodrugs, such as ethionamide, which need activation by normal metabolism to release their toxic effects. Ethionamide is potentiated by small molecules. Blondiaux et al. screened for more potent analogs and identified a lead compound called SMARt-420. This small molecule inactivates a TetR-like repressor, EthR2, and boosts ethionamide activation. SMARt-420 successfully promoted clearance of multidrug-resistant strains of Mycobacterium tuberculosis from the lungs of mice.

Science, this issue p. 1206

Abstract

Antibiotic resistance is one of the biggest threats to human health globally. Alarmingly, multidrug-resistant and extensively drug-resistant Mycobacterium tuberculosis have now spread worldwide. Some key antituberculosis antibiotics are prodrugs, for which resistance mechanisms are mainly driven by mutations in the bacterial enzymatic pathway required for their bioactivation. We have developed drug-like molecules that activate a cryptic alternative bioactivation pathway of ethionamide in M. tuberculosis, circumventing the classic activation pathway in which resistance mutations have now been observed. The first-of-its-kind molecule, named SMARt-420 (Small Molecule Aborting Resistance), not only fully reverses ethionamide-acquired resistance and clears ethionamide-resistant infection in mice, it also increases the basal sensitivity of bacteria to ethionamide.

Antibiotic resistance is a rapidly growing health concern and is observed for many antibacterial agents, both in hospital and in community settings (1, 2). The development of drug resistance—especially rifampicin resistance (RR), multidrug resistance (MDR) and extensive drug resistance (XDR)—is particularly worrisome for tuberculosis (TB) (3). Approximately 580,000 MDR/RR-TB cases have occurred in 2015, resulting in about 250,000 deaths. This situation seriously undermines efforts to control the global epidemic of TB and may soon counteract the slow but continuous annual decline of ~1.5% observed during the past 14 years (4).

Discovering new anti-TB therapeutics is difficult (5), and few new drugs have emerged during the past 30 years. Moreover, because current TB treatment requires poly-therapeutic approaches, losing key antibiotics because of the emergence of drug resistance may impair the efficacy of the whole combination.

Drug resistance in bacteria can occur through mutations in the antibiotic’s target; the acquisition of enzymes that modify or degrade the drug, such as aminoglycoside-modifying enzymes or β-lactamases; their active expulsion from the bacteria; or alterations of the cell permeability (6). Sometimes, antibiotic resistance can be reversed—for example, through restoration of the antimicrobial activity of β-lactams by clavulanic acid, which inhibits enzymes responsible for their degradation (7). Unfortunately, there are so far no other examples of this 40-year-old paradigm.

Some of the most effective anti-TB antibotics require bioactivation by Mycobacterium tuberculosis enzymes to acquire their antibacterial effect. These pro-antibiotics not only include the 40-year-old compounds isoniazid (INH), pyrazinamide (PZA), p-aminosalicylic acid (PAS) and ethionamide (ETH), but also the recently approved drug delamanid (OPC-67683) and the under-clinical-evaluation compound pretomanid (PA824). However, bioactivation of pro-antibiotics is vulnerable to mutational inactivation or attenuation of the corresponding bioactivating enzymes, as observed for INH-, PZA-, and ETH-resistant clinical isolates with mutations in katG (8), pncA (9), and ethA (10, 11), respectively. Similarly, experimentally generated and clinical resistance to delamanid and to pretomanid pointed to enzymes and coenzymes involved in their bioactivation (1214). Resistance to PAS also involves mutations in enzymes, such as dihydrofolate synthase, which is implicated in its bioactivation (15).

We have discovered a spiroisoxazoline family of Small Molecules Aborting Resistance (SMARt) that induces expression of an alternative bioactivation pathway of ETH, reverting acquired resistance of M. tuberculosis to this antibiotic.

The bioactivation of ETH in M. tuberculosis is normally catalyzed by the Baeyer-Villiger monooxygenase EthA (10, 11, 16). Transformation of ETH by EthA into highly reactive intermediates leads to the formation of a stable covalent adduct of ETH and nicotinamide adenine dinucleotide (NAD) (10, 17). This adduct binds to and inhibits the enoyl reductase InhA involved in mycolic acid biosynthesis, one of the essential components in the mycobacterial cell wall (18, 19). The production of EthA is regulated by the TetR-type transcriptional repressor EthR (20). Previously, we have shown that small-molecule inhibitors of EthR stimulate the transcription of the ethA gene (2124), which improves the bioactivation of ETH and consequently boosts its antibiotic activity, both in vitro and in vivo (25). These booster molecules, such as BDM41906, reduce or reset the innate resistance of M. tuberculosis to ETH; however, as expected, they were unable to boost the bioactivation of ETH in strains harboring mutations in ethA (Table 1, panel B).

Table 1 Impact of BDM41906 and SMARt-420 on the ethionamide susceptibility of a selection of clinical strains.

(Panel A) Antibiotic profile. Threshold concentrations above which bacteria are considered clinically resistant are indicated. The drug-sensitivity status of each strain is reported; green indicates “under the threshold concentration,” and red indicates “above the threshold concentration.” Specifically, for ETH, MICs have been defined by MGIT960 and are reported (values in micrograms per milliliter). All selected strains except the reference pan-susceptible laboratory strain H37Rv (group 1) are multidrug-resistant (INH- and RIF-resistant). Group 2 includes ETH-sensitive strains. Group 3 contains ETH-resistant strains without mutation in ethA. Group 4 contains ETH-resistant strains mutated in ethA. (Panel B) MIC of ETH in the presence of 10 μM first-generation booster BDM41906. (Panel C) MIC of ETH in the presence of 10 μM SMARt-420.


Embedded Image
View this table:

During optimization of first-generation EthR inhibitors, most derivatives revealed a good correlation between binding to EthR and ETH-boosting activity against the bacteria (2124). However, unexpectedly, the replacement of the oxadiazole-piperidine motif by a more constrained, structurally divergent, spiroisoxazoline scaffold completely abolished the ability of the compounds to bind to EthR in vitro, whereas they remained highly effective in boosting ETH activity against M. tuberculosis (Fig. 1A).

Fig. 1 SMARt family of molecules reveals alternative ETH bioactivation pathway.

(A) Bidimensional representation of the properties of ETH-boosting compounds (BDM). The x axis indicates the shift in the melting temperature (ΔTm) of EthR in the presence of BDM compounds, which translates the capacity of the compounds to bind and thermostabilize EthR in vitro (values are provided in table S1). The y axis indicates the potency [expressed as the negative logarithm of the median effective concentration (EC50)] of a panel of compounds to increase ETH antibacterial activity on M. tuberculosis–infected macrophages. EC50 is the concentration of compound that allows ETH at 0.1 μg/mL (10 times less than the normal MIC) to inhibit 50% of M. tuberculosis growth in macrophages. Blue dots and red dots represent compounds of the oxadiazole-piperidine family (first-generation boosters) and of the spiroisoxazoline family (SMARt), respectively. (B) RNA-seq analysis of genes that are differentially expressed in M. bovis BCG exposed for 24 hours to 25 μM BDM41906 (2) or SMARt-420 (3) in comparison with dimethyl sulfoxide–treated bacteria (1). Only genes showing a minimum twofold change (FC) in transcript abundance in at least one condition are shown. RPKM, reads per kilobase per million mapped reads. Whereas BDM41906 specifically induce the expression of ethA and ethR, SMARt-420 massively induces the expression of bcg_0108c and bcg_0107c. A weak, but statistically significant, induction of ethA and ethR is also observed. (C) Comparison of the genetic organization and predicted function of the corresponding M. tuberculosis loci and proteins. Genes bcg_0107c and bcg_108c correspond to rv0076c and rv0077c, respectively. Rv0078 is predicted as a transcriptional repressor of the TetR family.

Because these compounds had no antibacterial activity in the absence of ETH but boosted ETH activity independently of EthR, we hypothesized that they may trigger an alternative bioactivation pathway for ETH.

To identify this pathway, we studied the impact of SMARt-420, a representative member of the spiroisoxazoline family, on the transcriptome of Mycobacterium bovis Bacillus Calmette-Guérin (BCG) and compared it with the impact of the bona fide EthR-inhibitor BDM41906 (23).

When the mycobacteria were treated with BDM41906, overexpression of both ethA and ethR was observed (Fig. 1B), which is in agreement with previous reports showing that EthR represses both ethA and its own expression (20). No other major modification of the transcriptome was observed, suggesting that the inhibitory activity of BDM41906 is restricted to EthR. In contrast to BDM41906, SMARt-420 only weakly induced the expression of ethA and ethR. However, SMARt-420 strongly activated the expression of the distantly located group of genes bcg_0107cbcg_0108c, corresponding to rv0076crv0077c (Fig. 1C) in M. tuberculosis. According to protein homology, Bcg_0108c is predicted to be a member of the large family of oxidoreductases (http://enzyme.expasy.org/EC/1.-.-.-), which also includes EthA. In silico analyses revealed that bcg_0108c is neighboring the tetR type transcriptional regulator gene bcg_0109 (http://pfam.xfam.org/family/PF00440), indicating analogies between the rv0076c-rv0078 and the ethR-ethA loci (rv3854c-rv3855). The genetic organization of the two loci is also similar: rv0077c and rv0078, like ethA and ethR, are divergent open reading frames, both separated by small intergenic regions [76 base pairs (bp) and 62 bp for ethR-ethA and rv0077c-rv0078, respectively] (Fig. 1C). By analogy with the transcriptional organization of the ethA-ethR regulon (20), these observations indicate that Rv0078 might regulate the expression of rv0077c by binding within the intergenic region. This hypothesis was confirmed with surface plasmon resonance experiments (SPR, Biacore), demonstrating the specific binding of Rv0078 to the intergenic region of rv0077c-rv0078 (Fig. 2A). In contrast, no binding of EthR to the intergenic region of rv0077crv0078 was observed, even at high concentrations of protein, thus excluding EthR for controlling the expression of the rv0076crv0078 locus (Fig. 2B). Conversely, no interaction was detected between Rv0078 and the ethAethR intergenic region, indicating the absence of cross-talk between the two regulons (Fig. 2B). We assigned the names EthR2 and EthA2 to Rv0078 and Rv0077c, respectively.

Fig. 2 SMARt-420 interacts with the transcriptional regulator EthR2.

(A) Global affinity of the interaction between EthR2 and the ethA2-ethR2 intergenic region measured with SPR. Sensorgrams of 22.7, 11.35, 5.60, 2.83, 1.41, and 0.28 nM EthR2 (Rv0078) injections over a sensor chip functionalized with 40 RU (resonance unit) of the biotinylated ethA2-ethR2 intergenic DNA. (B) Comparison of the injection of EthR and EthR2 (Rv0078) over 2 sensor chips functionalized with ethA-ethR intergenic DNA (top) and with ethA2-ethR2 intergenic DNA (bottom), respectively. (C) Binding of SMARt-420 and BDM41906 on EthR2 measured by determination of the melting temperature (Tm) of the complex (thermal-shift assay) (supplementary materials, materials and methods). (D) Crystal structure of the EthR2/SMARt-420 complex and illustration of the steric inability of binding of the repressor to DNA because the 40.7 Å spacing of the HTH motifs. The zoom onto the ligand binding pocket of the protein shows the an omit map contoured at 1.2σ for one of the two SMARt-420 molecules embedded in each monomer of the EthR2 homodimer (statistics are provided in the supplementary materials, materials and methods, and table S2).

The direct binding of SMARt-420 to EthR2 was analyzed in vitro by means of thermal shift assay. The dose-dependent thermal stabilization of EthR2 through binding to SMARt-420 is illustrated in Fig. 2C, whereas no interaction between BDM41906 and EthR2 was observed at equivalent concentrations, which is in agreement with the lack of effect of BDM41906 on the transcription of ethA2.

A detailed understanding of the interaction of SMARt-420 with EthR2 was obtained from the x-ray structure of the complex, which shows that EthR2 forms a homodimer in which one molecule of SMARt-420 is embedded in each monomer (Fig. 2D). The structure also confirmed EthR2 as a typical TetR-type regulator harboring two twistable helix-turn-helix (HTH) motifs that are typically involved in the binding of the homodimer to its DNA target (26). In this family of repressors, binding of ligands to the distant specific pockets located in the regulatory core of the homodimer induces allosteric reorganization of the architecture of the HTH motifs and thereby modifies the binding properties of the protein to its DNA target (23). In agreement with this paradigm, the EthR2–SMARt-420 cocrystal revealed that the distance separating the HTH motifs is far larger than the ±34 Å required for the binding of the regulator to DNA (2729), thus providing the mechanism of action of SMARt-420 (Fig. 2D).

To quantify the inhibition of EthR2-binding to its DNA target by SMARt-420, we designed a synthetic mammalian gene circuit that senses the EthR2-DNA interaction in human cells and produces a quantitative reporter gene expression readout [secreted alkaline phosphatase (SEAP)] (Fig. 3A) (30). In contrast to its repressor role in mycobacteria, binding of EthR2 to the chimeric promoter used in this assay is expected to activate the expression of the SEAP reporter gene. In the absence of SMARt-420, we observed strong SEAP production, confirming the binding of EthR2 to its DNA promoter in this cellular context. Upon adding SMARt-420, a dose-dependent inhibition of SEAP production was observed, confirming that SMARt-420 impairs the DNA-binding properties of EthR2. In contrast and as expected, no effect was observed when the cells were incubated with BDM41906, confirming the specificity of the SMARt-420–EthR2 interaction (Fig. 3A). In vitro, SPR experiments showed that SMARt-420 inhibits in a dose-dependent manner the binding of EthR2 to its DNA target (Fig. 3B), thus confirming that no other partner is required.

Fig. 3 SMARt-420 inhibits the DNA binding activity of EthR2.

(A) Synthetic mammalian gene circuit designed to sense EthR2-DNA interactions (ethA2ethR2 intergenic region) that are required to induce expression of the reporter gene SEAP. The alkaline phosphatase induced by the EthR2-VP16 complex was inhibited in a dose-dependent manner by SMARt-420, whereas no effect was observed in the presence of BDM41906 (supplementary materials, materials and methods, and fig. S1). (B) Inhibition of the binding of EthR2 to the ethA2ethR2 intergenic region by SMARt-420 but not by BDM41906, measured with surface plasmon resonance. (C) Effect of the overexpression of ethA2 (pMV261–ethA2) on the MIC of ETH in M. tuberculosis.

The expression of ethA2 upon inhibition of EthR2 by SMARt-420 leads to efficient bioactivation of ETH in the bacteria. To evaluate the role of EthA2 in the SMARt-420–controlled bioactivation of ETH, M. tuberculosis H37Rv was engineered to overexpress ethA2 by using the multicopy plasmid pMV261 (31). Under these conditions, the minimal inhibitory concentration (MIC) of ETH decreased from 2 to 0.25 μg/ml, suggesting that EthA2 takes part in the bioactivation of ETH when overexpressed, thus reducing the innate resistance of the bacteria to ETH (Fig. 3C).

SMARt-420 is the most active compound of a spiroisoxazoline series for which the binding to EthR2 (thermal shift assay) and the inhibition of the DNA binding of EthR2 (SEAP assay) were shown to be correlated to the ETH-boosting effect on sensitive and on ETH-resistant M. tuberculosis (fig. S2).

As indicated by the transcriptomic analyses, the basal expression level of ethA2 in the absence of SMARt-420 is low in M. tuberculosis. We measured the relative abundance of mRNA by means of high-throughput RNA-sequencing (RNA-seq) in M. tuberculosis cells grown to mid-log phase. We found that ethA2 belongs to the 10% least-expressed genes in M. tuberculosis (rank 3609 out of the 3973 TB genes), whereas ethA is among the 10% genes with the highest levels of expression. When compared with each other, ethA2-mRNAs were about 60 times less abundant than ethA-mRNAs in M. tuberculosis grown under standard conditions. Low expression of ethA2 in the absence of EthR2 inhibitors is of clinical importance, offering an explanation as to why EthA2 has not been previously identified as involved in ETH activation, and therefore mutations in this gene have not been observed in clinical isolates resistant to ETH so far.

Human-adapted M. tuberculosis complex comprises seven main phylogenetically distinct lineages (32). We analyzed the genome sequences of 217 geographically diverse clinical strains representing all seven lineages (33) and confirmed the presence of the rv0076crv0078 locus in all strains. In addition, the presence of the locus was confirmed in a collection of ETH-resistant (171) and ETH-sensitive (253) clinical isolates of M. tuberculosis (34). No mutation in this locus was observed in the ETH-sensitive population. Only one ETH-resistant isolate shows a point mutation in rv0078, which also contains a Stop mutation in ethA, which is most probably responsible for the ETH-resistance phenotype (table S3). Then, a panel of ETH-sensitive and ETH-resistant MDR M. tuberculosis clinical isolates were tested by means of respirometry (MGIT 960) (35) for their sensitivity to ETH in the presence of SMARt-420. Treatment of ETH-sensitive strains with SMARt-420 (10 μM) decreased their ETH MIC (Table 1, panel C). Strains highly resistant to ETH, because of mutations in ethA, were also sensitive to ETH in the presence of SMARt-420 (Table 1, panel C, group 4). The combination SMARt-420 plus ETH was active against all ETH-resistant, MDR, and XDR isolates tested. Last, we showed that overexpression of inhA by using a multicopy plasmid modified the boosting effect of SMARt-420, suggesting that the new bioactivation pathway of ETH still targets InhA (fig. S3). We also verified that SMARt-420 does not affect the efficacy of other antibiotics (table S4).

Pharmacokinetic experiments performed in female Swiss mice showed that a single oral dose of 30 mg/kg of SMARt-420 provides a circulating concentration of SMARt-420 higher than that required to boost ETH in vitro (table S5). Restoration of sensitivity to ETH by SMARt-420 was evaluated in C57BL6/J mice infected by aerosol with 105 ETH-resistant M. tuberculosis bacilli mutated in ethA. Seven days after infection, the mice were treated with ETH alone or with ETH in combination with SMARt-420. Daily administration of up to 50 mg/kg of ETH for 3 weeks was ineffective in significantly reducing the bacterial load in the lungs (Fig. 4), confirming the resistance to ETH of this strain. In contrast, mice treated with a combination of ETH and SMARt-420 (both at 50 mg/kg) showed a striking reduction of the bacterial load (4.6 log) in the lungs (Fig. 4). The absence of an effect observed with SMARt-420 administered alone, and a dose-response to ETH combined with SMARt-420, confirms that the anti-TB activity of the combination is specifically due to the restoration of ETH sensitivity of this strain.

Fig. 4 Reversion of ETH resistance in tuberculosis-infected mice.

Mice (five mice per group) infected with ETH-resistant bacteria were treated with the control antibiotic INH (25 mg/kg), ETH alone (50 mg/kg), SMARt-420 alone (50 mg/kg), or a combination of ETH and SMARt-420. Pulmonary bacillary loads were enumerated by colony-forming units (CFUs) after 3 weeks of treatment. Administration of up to 50 mg/kg of ETH did not reduce the pulmonary load of ETH-resistant mycobacteria, whereas coadministration of ETH and SMARt-420 showed a dose-dependent reduction, with a maximum of 4.6 log (control versus ETH50+SMARt-420). Details and statistics are available in “Bonferroni’s multiple comparison test” in table S6.

We show that drug-resistance to the widely used antituberculosis drug ETH (36) can be circumvented by spiroisoxazoline SMARt-420, which activates cryptic drug-bioactivation pathways in drug-resistant pathogens. Other ETH activation pathways may exist in M. tuberculosis, including the recently described VirS-MymA (37), opening supplementary avenues for reversing ETH-resistance and boosting its activity.

Innovative treatment protocols could also be explored, in which noncontinuous but regular administration of SMARts to TB patients would periodically toggle the expression of alternative bioactivation pathways of pro-drugs. This approach could be used to limit the frequency of resistance by systematically destroying subpopulations of resistant bacteria that may emerge during treatment.

Supplementary Material

www.sciencemag.org/content/355/6330/1206/suppl/DC1

Materials and Methods

Figs. S1 to S5

Tables S1 to S6

References (3845)

Database S1

References and Notes

  1. Acknowledgments: All data and code to understand and assess the conclusions of this research are available in the main text, supplementary materials, and via the following repositories: The refined coordinates and the structure factors of Rv0078 were deposited in the Protein Data Bank under the accession numbers 5N1C (iodinated form), 5N1I (unliganded form), and 5ICJ (liganded form); raw data of RNA-seq analysis have been deposited in Datadryad.org under the doi 10.5061/dryad.mb463. We are indebted to the Soleil [Block Allocation Group (BAG) proposal 20141408] and the European Synchrotron Radiation Facility (BAG proposal MX-1677) synchrotrons for beam-time allocations on this project. Sequencing analyses were performed at the sciCORE (http://scicore.unibas.ch) scientific computing core facility at the University of Basel. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. We are indebted to E. Willery for technical support in molecular biology, F. Leroux for screening management, L. Agouridas and N. Probst for chemical synthesis, and C. Piveteau for bioanalysis. This work was supported by l’Agence Nationale de la Recherche (ANR), France (Tea-4-Two, ANR-14-CE14-0027-01) (ANR-10-EQPX-04-01), by EU grants ERC-STG INTRACELLTB no 260901, the Feder (12001407 (D-AL), PRIM (NewBio4Tb), European Research Council (grant 309540-EVODRTB), SystemsX.ch, Institut National de la Santé et de la Recherche Médicale, Université de Lille, Institut Pasteur de Lille, Centre National de la Recherche Scientifique, the Région Hauts-de-France (convention no. 12000080), and Société d’Accélération du Transfert de Technologie Nord. R.W. is Research Associate at the National Fund for Scientific Research (FNRS-FRS) (Belgium). M.M. was supported by PRIM (NewBio4Tb), V.D. was supported by EU grant 260872, and V.T. was suppported by the Marie Curie Initial Training Network (ITN-2013-607694-Translocation). The authors also thank the Unité Mixte de Recherche UMR 8199, Lille Integrated Genomics Network for Advanced Personalized Medicine (LIGAN-PM) Genomics platform (Lille, France), which belongs to the Federation de Recherche 3508 Labex EGID (European Genomics Institute for Diabetes; ANR-10-LABX-46) and was supported by the Agence Nationale de la Recherche (ANR) Equipex 2010 session (ANR-10-EQPX-07-01; LIGAN-PM). The LIGAN-PM Genomics platform is supported by the Fonds Européen de Développement Régional and the Region Hauts-de-France. We thank A. Wagner and Roquettes-Frères (Lestrem, France) for their gift of Cyclodextrine Kleptose HP. We greatly appreciate the fruitful discussions with all the members of the GlaxoSmithKline team “Diseases of the Developing World (DDW)” and their invaluable support in the continuation of this project. C.K. is employed at Bioversys; S.G. is a consultant for the Foundation of Innovative Diagnostics (FIND) in Geneva, Switzerland. M.G. is CEO of Bioversys and a Board Member of the BEAM Alliance, a group of Biopharmaceutical companies from Europe innovating in antimicrobial resistance research. The BEAM Alliance is a not-for-profit association, and there is no financial remuneration of any kind for its Board members. All companies included are working within the field of antimicrobial drug discovery. N.W., B.D., A.R.B., P.B., and M.D. are inventors on patent PCT/EP2013/077706, which covers BDM41420; WO/2014/096369. B.D., N.W., C.L., and A.R.B. are inventors on patent PCT/FR2007/001138; WO/2008/003861, which covers BDM41906. The facilities conformed to Directive 86/609/EEC on the Protection of Animals Used for Experimental and Other Scientific Purposes and norms published in the European Council ETS123 Appendix A. Facilities and procedures complied with the Belgian Law of 14 August 1986 on animal protection and welfare. Training of experimental leaders, biotechnicians, and animal caretakers was in accordance with Royal decree of 13 September 2004, which specifies the training of persons working with laboratory animals. All animal experimentation and procedures performed at the National Reference Center for Tuberculosis and Mycobacteria, Bacterial Diseases Service, Scientific Institute of Public Health (WIV-ISP), Brussels, Belgium, were validated and approved by the Ethical Committee of the IPH-VAR (Scientific Institute of Public Health–Veterinary and Agrochemical Research Centre, Belgium) under file no. 120323-01. The animal facilities and procedures were under the supervision of an expert on animal welfare in accordance with the Belgian Ministry of Health. pET-15b-ethR2 and pET-15b-ethR are available from A.R.B. under a materials transfer agreement with Institut Pasteur de Lille. pCK289 and pCK287 are available from M.G. under a materials transfer agreement with Bioversys.
View Abstract

Related Content

Subjects

Navigate This Article